The Medical Dictionary for Regulatory Activities (MedDRA) is the new International Conference on Harmonisation (ICH) approved global standard terminology for medical data intended to be used for all regulatory activities. While MedDRA is currently used primarily in pharmacovigilance, its use in clinical development and regulatory affairs is rapidly increasing. The terminology is now mandatory for the coding of adverse drug reactions (ADRs) and adverse events (AEs) in the submission of expedited reports, periodic safety update reports and for the 6-year re-examination after approval to the Ministry of Health and Welfare (MHW) in Japan and it is also likely to become mandatory in the EU and the US. Together with E2B and M2 standards, MedDRA is one of the key elements for electronic transmission of safety data. Electronic transmission of individual case reports does not work without the use of this new terminology.

MedDRA is an extensive and complex terminology with more than 60 000 terms. In version 6.0, 58 560 terms are intended for the term selection for coding. These so-called ‘Lowest Level Terms’ (LLTs) allow a very high degree of specificity at the data entry level. This supports accuracy and precision in medical coding. Interpretation prior to coding — a procedure often necessary with the previous terminologies because of lack of specificity — is superfluous with MedDRA; in fact, interpretation should be avoided at this point. Unambiguous data and correct data input are essential preconditions for meaningful data retrieval, analysis and medical evaluation. For optimal use of MedDRA, an ICH-endorsed “MedDRA points to consider” document is available for term selection in the newest version of the Maintenance Supply and Support Organisation (MSSO) website.[1]

Throughout the development of MedDRA, the ICH Expert Working Group (M1) has tried to balance specificity of coding with creating a structure that facilitates data retrieval and analysis. The result was a five-level hierarchy with multiaxial relationships completed by so-called ‘Special Search Categories’ (SSCs). Currently, a more sophisticated concept for data analysis is under development in a joint initiative of the Council for International Organizations of Medical Sciences (CIOMS) and MSSO, the so-called ‘Standardised MedDRA Queries’ (SMQs).[2] These retrieval tools are aimed at identifying all cases in a database that are related to the diagnoses and syndromes most relevant in pharmacovigilance. Although MedDRA has been primarily developed in the context of pharmacovigilance, the full implementation of MedDRA should also cover its use in clinical trials, as pointed out in the ICH M4E: Common Technical Document — Efficacy.[3] See Kübler et al.[4] for an in-depth discussion of issues relating to this field and proposals for standard tables.

In order to give the reader a basis for the understanding of how retrieval strategies may need to change with MedDRA, the structure and some key features of MedDRA will be briefly explained. Next, two examples of MedDRA-coded ADR data, one from a clinical study project and one from spontaneous reporting, are presented. The first example compares the ADR profile of the same data coded in WHO Adverse Reaction Terminology (WHO-ART) and re-coded in MedDRA. The second example demonstrates different ways of presenting the ADR profile of an antibacterial, using different elements in MedDRA with the aim of presenting the ADR profile in a way that best transports the medical concepts reported with this drug. At the end, critical findings from the demonstrated examples and from a review of the terminology triggered by work on queries and data analysis in daily pharmacovigilance are discussed with regard to their possible impact on data analysis and presentation.

Description of the Structural Elements and Key Features of the Medical Dictionary for Regulatory Activities (MedDRA)

The following description of MedDRA is taken from the descriptive documents released by the MSSO as part of the license, and from the examples found by browsing the dictionary using the MSSO browser, which is released with the terminology.

Five-Level Hierarchy

The MedDRA hierarchy is shown in figure 1. The 16 341 Preferred Terms (PTs) represent single medical concepts clearly distinguished from one another. PTs should be unambiguous, specific and self-descriptive. The single medical concepts at the PT level are very specific to allow a high degree of medical correctness and precision in coding. Examples of PTs include ‘Abdominal pain upper’, ‘Abdominal pain lower’, ‘Abdominal tenderness’, ‘Abdominal rebound tenderness’, ‘Acute abdomen’, etc. The 58 560 Lowest Level Terms (LLTs) below the PTs reflect the same medical concept as the corresponding PT but include relevant synonyms and lexical variants. For example, the PT ‘Abdominal pain’ groups LLTs such as ‘Abd. pain’, ‘Abdominal pain’, ‘Abdominal cramps’, ‘Abdominal wall strained’, etc. The LLTs, therefore, allow the preservation of the original information to a larger degree than, for example, COSTART, where only the term ‘Abdominal pain’ is available.

Fig. 1
figure 1

Five-level hierarchy (Medical Dictionary for Regulatory Activities [MedDRA] version 6.0).

Full size image

High Level Terms (HLTs) have been created to group the specific PTs under broader medical concepts using different classifications, such as anatomy, physiology, pathology, mechanisms, etc. as appropriate in order to allow aggregation of data. All PTs have fix links to at least one HLT. For example, the HLT ‘Duodenal ulcers and perforation’ groups nine PTs describing different forms of ulcers and perforated ulcers.

The next level are the High Level Group Terms (HLGTs), which group the HLTs again by a broader concept and can also combine different classifications, such as anatomy, pathophysiology, etc. For example, the HLGT ‘Gastrointestinal ulceration and perforation’ groups seven HLTs by anatomy and mechanism.

The highest level of the hierarchy is the 26 System Organ Classes (SOCs), which is the same concept as the ‘Body Systems’ that we know from previous terminologies. However, the following five SOCs cover new areas:

  • infections and infestations

  • injury, poisoning and procedural complications

  • surgical and medical procedures

  • investigations

  • social circumstances.

MedDRA has not just been developed as a new thesaurus for AEs — it has a much broader scope, i.e. it is supposed to be used for signs and symptoms, diseases, diagnoses, therapeutic indications, names and qualitative results of indications, surgical and medical procedures, and medical, social and family history. Some of the above-mentioned SOCs are clearly related to these additional fields. Currently, MedDRA provides no guidance as to which terms should be used for which purpose of retrieval and presentation, and this may impact the retrieval, analysis and presentation strategy.

Multiaxiality

Each PT is represented in at least one SOC. When a PT is allocated to more than one SOC, one SOC is defined as the primary SOC. Assignment of the primary SOC simply follows rules, but does not imply medical judgement or priority over the other SOCs. This concept intends to generate standardised presentation of data by primary SOC and to count each PT only once in its primary SOC. For example, the PT ‘Vision blurred’ is allocated to the SOC ‘Eye disorders’ and to the SOC ‘Nervous system disorders’. The primary SOC is ‘Eye disorders’, following the rule that the prime manifestation site usually defines the primary SOC.

There are several exceptions to this rule and it is very important for data analysis to understand that the following three SOCs are ‘stand alone’:

  • surgical and medical procedures

  • investigations

  • social circumstances.

This means that terms are only represented in these SOCs without having secondary links to other SOCs. As an important consequence for data analysis, the SOC ‘Investigations’ must always be searched in addition to the ‘Disorders’ SOC. For example, the results of liver function tests are not represented under the SOC ‘Hepatobiliary disorders’ but under the SOC ‘Investigations’.

Special Search Categories (SSCs) and Standardized MedDRA Queries (SMQs)

As with any other terminology, some clinical conditions and syndromes cannot be fully represented in one SOC, such as ‘anaphylaxis’ or ‘haemorrhage’. This is because they affect various parts of the body at the same time and the related terms are spread over more than one SOC and cannot be completely grouped by HLT or HLGT. For such searches, MedDRA originally proposed the concept of ‘Special Search Categories’ (SSCs), where all PTs relevant for a certain condition are grouped together across SOCs.

Currently, the SSCs include only PTs. Combination of PTs and HLTs/HLGTs is not possible and no hierarchy of term selection is proposed. Therefore, in a co-operative effort between a CIOMS working group and the MSSO of MedDRA, the concept of ‘Standardized MedDRA Queries’ (SMQs) has been conceived to allow these combinations and includes a hierarchical approach and algorithms for combination of relevant terms. There is an intention to cover the most frequent and relevant issues in pharmacovigilance.[2] SMQs will become part of MedDRA, as they will be finalised, tested and maintained by the MSSO.

Practical Examples of Data Presentation of MedDRA-Encoded Data

Two sets of data are presented. The first example is taken from a clinical study project, and the second example is taken from a database of spontaneously reported cases. The first example is chosen to demonstrate if and how the safety profiles look different when data that were initially coded in WHO-ART were re-coded in MedDRA. The second example describes different ways of sorting spontaneously reported safety data of an antibacterial, and focuses on how to best present the data to enable a reviewer to quickly understand the most frequently reported medical concepts of the safety profile.

Comparison of Adverse Drug Reaction Profiles by WHO Adverse Reaction Terminology (WHO-ART) and MedDRA

WHO-ART-encoded safety data from 15 clinical studies of the same project, including 847 patients in the verum group and 127 patients in the placebo group, were re-coded with MedDRA version 2.0 on the basis of the verbatim terms. The coded data were checked for consistency and medical accuracy according to pre-fixed company coding guidelines, based upon MedDRA rules and conventions and the “MedDRA term selection: points to consider”.[1] The results are presented in table I and figure 2.

Table I
figure Tab1

Number of distinct categories for eventsa

Full size image
Fig. 2
figure 2

Adverse drug reaction profile of verum group: distribution of System Organ Classes (SOCs) using WHO Adverse Reaction Terminology (WHO-ART) or Medical Dictionary for Regulatory Activities (MedDRA) [only SOC/body class >5% are presented]. GI = gastrointestinal.

Full size image

Table I shows the effect of the greater granularity in MedDRA. In WHO-ART, 214 different PTs are used; whereas, in MedDRA, 321 different PTs describe the reported terms much more specifically. Figure 2 shows that many laboratory terms that show up in WHO-ART under Body Systems ‘Metabolic/nutritional disorders’, ‘Liver disorders’ and ‘Other’ are grouped together under the SOC ‘Investigations’ in MedDRA. This leads to ‘Liver disorders’ disappearing from the list of SOCs that contain more than 5% of the PTs. Conversely, the SOC ‘Infections and infestations’ — a new category in MedDRA — is populated with 6% of the PTs, whereas this concept is not represented by WHO-ART at this level. It should be noted that the percentages must not be interpreted as incidence rates, since they are calculated as the number of events coded to the respective SOC divided by the total number of events.

In summary, based upon the same verbatim information, there are 1.5 times more coded terms in MedDRA, and they are grouped quite differently in MedDRA compared with WHO-ART. This result is in line with findings of other authors.[5]

Presentation of Spontaneous Reports Comparing Different Approaches to Sorting the Data

Spontaneously reported safety data of an antibacterial covering a 10-year period, including 1229 case reports with 2174 ADRs, were coded with MedDRA version 5.0. The data are stored in Clintrace 2.8™ Footnote 1. For analysis, the key variables (suspect drug, demographics, case ID, clinical event, etc.) were exported from Clintrace via SAS/ACCES® interface to Open DataBase Connectivity (ODBC), a validated tool of SAS®. The data were then merged with the MedDRA files and analysed using validated SAS® tools.

The purpose of this example is to illustrate how a medically meaningful AE profile can be derived based on MedDRA coded data. For the sake of simplicity, percentages are the number of ADRs divided by the total number of ADRs, instead of number of cases with specific ADRs divided by number of cases.

The effects of different ways of data presentation on clarity and medical meaningfulness of the ADR profile were checked by drug safety physicians for plausibility by listings of the cases retrieved under the different strategies.

In the first step, the data were broken down in a standard way by primary SOC and PT. Further steps included all levels of the MedDRA hierarchy as it is and a free combination of structural elements by medical concepts.

The analysis by primary SOC shows that 22 different SOCs are populated with terms. This is illustrated in figure 3. Only six SOCs contain at least 5% of the PTs, as exhibited in figure 4.

Fig. 3
figure 3

Adverse drug reaction (ADR) profile by primary System Organ Class.

Full size image
Fig. 4
figure 4

Primary System Organ Classes containing >5% of preferred terms. ADR = adverse drug reaction; infec = infections and infestations; gastr = gastrointestinal disorders; genrl = general; nerv = nervous system disorders.

Full size image

In the next step, the group terms of the MedDRA hierarchy, i.e. HLGT and HLT, were included to exhibit the most relevant sub-groups of terms in the six SOCs, which contain at least 5% of the PT. We choose the most populated HLGTs in the respective SOC (which contain >1% of the total of reported PTs). This is displayed in figure 5.

Fig. 5
figure 5

Primary System Organ Classes (SOCs) [>5% of total Preferred Terms] substructured by High Level Group Terms (HLGTs) [>1% of total Preferred Terms]. ADR = adverse drug reaction; gastr = gastrointestinal disorders; genrl = general; GI = gastrointestinal; infec = infections and infestations; NEC = not elsewhere classified; nerv = nervous system disorders.

Full size image

Next, each of the six SOCs containing at least 5% of the PTs were broken down by displaying the HLGTs >1% and under these, the most populated HLTs (containing at least 1% of the total of reported PTs). The example of the SOC ‘Skin and subcutaneous tissue disorders’ is shown in figure 6.

Fig. 6
figure 6

System Organ Class ‘Skin and subcutaneous tissue disorders’ broken down by High Level Group Terms (HLGTs) and High Level Terms (HLTs) [>1% of total Preferred Terms]. ADR = adverse drug reaction; NEC = not elsewhere classified.

Full size image

The ‘split of medical concepts’, which is mainly caused by the rules and conventions related to multiaxiality (e.g. terms related to infections are displayed in the primary SOC ‘Infections and infestations’) versus ‘stand alone SOCs’ (e.g. SOC ‘Investigations’) leads to the situation where the ADR profile for this antibacterial — gained by the standard display by primary SOC — exhibits nearly 15% of the PTs under the SOC ‘Infections and infestations’ (see figure 4). The laboratory terms, e.g. related to hepatic disorders, are classified under the SOC ‘Investigations’ and not — as medically more meaningful — under the SOC ‘Hepatobiliary disorders’.

Nearly 15% of terms were placed under the SOC ‘Infections and infestations’ because reports related to diarrhoeas caused by clostridium difficile are classified to the infection as the primary allocation and not to the colitis (classified under the SOC ‘Gastrointestinal disorders’).

Similarly, relevant terms related to other medical concepts, such as allergic reactions, are found under the SOC ‘Immune system disorders’ and do not all appear under the SOC ‘Skin and subcutaneous tissue disorders’, which is the primary manifestation site for allergies associated with this antibacterial.

One can say that the standard display by primary SOCs underestimates the real frequency of some medical concepts, such as colitis or allergy, given the fact that the PTs belonging to these concepts are not all grouped together in one SOC, or under one HLGT or HLT. Therefore, they are not perceived as belonging together in the standard print-out.

In order to reflect the medical conditions in a more meaningful way, deviating from the defined structure of MedDRA, the data were organised following medical considerations in the following way: primary SOCs were completed by HLGTs and/or HLTs from other SOCs as medically appropriate. To avoid double counting, the respective HLGTs or HLTs were removed from their primary SOCs.

For example, the primary SOC ‘Gastrointestinal disorders’ was completed with the HLT ‘Clostridia infections’ from the SOC ‘Infections and infestations’ and the HLGT ‘Gastrointestinal investigations’ from the SOC ‘Investigations’. Both HLT and HLGT were subtracted from their original primary SOC. By doing so, all terms related to the concept of pseudomembraneous colitis or diarrhoea associated with clostridium difficile were grouped together under the SOC of the primary manifestation site, which is the SOC ‘Gastrointestinal disorders’. Similarly, all HLGTs in the SOC ‘Investigations’ that clearly could be conceptually linked to a disorder SOC were linked to the respective disorder SOC and subtracted from the SOC ‘Investigations’. For example, the HLGT ‘Hepatobiliary investigations’ was removed and linked to the SOC ‘Hepatobiliary disorders’. The result is shown in figure 7.

Fig. 7
figure 7

Primary System Organ Classes (SOCs) containing >5% of Preferred Terms and including relevant High Level Group Terms (HLGTs) from other SOCs. ADR = adverse drug reaction; gastr = gastrointestinal disorders; genrl = general; hepat = hepatobiliary disorders; infec = infection; inv = investigation; nerv = nervous system disorders.

Full size image

Clearly, there is an important difference with regard to the percentage of PTs represented under the respective SOC and with regard to the fact that one SOC (‘Infections and infestations’) has disappeared and another SOC (‘hepatobiliary disorders’) has been added, because it contains now more than 5% of the total PTs. Details of the shift in the percentages of PT per SOC are illustrated in figure 8.

Fig. 8
figure 8

Transition from primary System Organ Class (SOC) to modified presentation by primary SOC >5% including relevant High Level Group Terms (HLGTs) from other SOCs. Gastr = gastrointestinal disorders; genrl = general disorders; hepat = hepatobiliary disorders; infec = infections and infestations; inv = investigations; nerv = nervous system disorders.

Full size image

Critical Findings and Discussion

In the following sections, some critical findings are discussed, which were found by the authors when working with MedDRA for data analysis and presentation. They reflect more findings than demonstrated in the examples under section 2 and are judged by the authors as having significant influence on data output in general.

The Preferred Term Level

In the data set of spontaneously reported cases of suspected ADRs demonstrated in section 2.2, the objective was mainly to find out how to best sort the data in order to present them in a way that most accurately describes the ADR profile for an assessor who is not yet familiar with the drug and its ADR profile. This approach attempts to group the data in a way that allows the user to quickly focus on the most relevant medical concepts and issues. A leading question was whether a standard approach of displaying the data by SOC and PT would be enough and whether additional approaches would add value by presenting the data more accurately from a medical perspective?

The standard analysis of ADR data includes presentation by SOCs and PTs. Presentation by SOCs only provided a very rough overview on the safety profile. Looking at the PT level hardly gave a clear overview either, because in the example of spontaneous data, 444 different PT were reported, covering more than 12 pages in a line listing. From these listings, it was difficult to recognise medically meaningful groupings of terms. When the PTs were grouped by frequency, only 15 PTs out of 444 (3.38%) reached frequencies >1% of the total number of PTs. While this approach might yield the most relevant PTs one cannot be sure that other medically related PTs are not overlooked.

As a consequence of high specificity, the number of reported events per specific PT was very small for the vast majority of PTs. This effect is in line with observations of other authors.[6,7] Clinically related PTs may be not recognised as belonging together. For clinical study programmes, this effect increases the danger of overlooking syndromes and differences in incidences of AEs/ADRs between placebo and verum. Statistical significance may not be reached, if the analysis is only done on the PT level. This also has a major impact on the frequency tables used for deciding on inclusion of terms in a Company Core Data Sheet (CCDS).

Similarly, the high granularity of MedDRA increases the risk of misclassification.[4,7–9] Minor difference in the verbatim may even lead to coding to completely different SOC. As an example, the PTs ‘Liver function test abnormal’ and ‘Function liver abnormal’ attribute to the SOCs ‘Investigations’ and ‘Hepatobiliary disorders’. This demonstrates that risk of misclassification is present at all hierarchy levels of MedDRA. If differences in ADR rates exist between different treatment regimens, misclassification can lead to the underestimation of relative risks,[9] i.e. decrease the chance to detect increased drug risks.

The MedDRA Group Terms (High Level Group Terms/High Level Terms)

As MedDRA claims that the group terms aggregate the data by meaningful medical concepts, the data were displayed by SOC and HLGT populated with >1% of the total number of PTs, as demonstrated in figure 5 and figure 6. Presenting the data this way substructures the SOCs by more specific concepts. However, the concepts at the HLGT level are often too broad to understand the data. Therefore, as demonstrated in figure 6, the SOC ‘Skin and subcutaneous tissue disorders’ was broken down by HLGT and HLT >1% of total PTs. The names of the HLTs are by far more understandable and a reviewer can quickly recognise the most relevant concepts under the displayed SOC. To a certain extent, a display by HLT overcomes limitations of some HLGT nominations that are not self-explanatory, and is a useful level to get an overview on clusters of PTs that are supposed to belong to the same medical concept. However, on both HLGT/HLT levels sometimes the nomination is not self-explanatory. For example, most of the terms with not elsewhere classified (NEC) are difficult, if not impossible, to understand without browsing the dictionary. Examples include the HLGT ‘Neurological disorders NEC’ or the HLT ‘Allergic condition NEC’ under an HLGT ‘Allergic conditions’.

For study analysis, it should be realised that significant differences between placebo and verum might appear, when group terms are included in the analysis, that group PTs by broader concepts. However, such groupings might be too broad with most of the HLGTs and sometimes not broad enough with some HLTs. Regrettably, the MedDRA HLGTs and HLTs do not represent a homogenous level of detail and can even vary significantly in their level of detail. The number of PTs under a HLT can vary from just one, for example, the HLT ‘Rabies viral infections’ with the only PT ‘Rabies’, to more than 80 PTs under the HLT ‘Cerebrospinal fluid tests (excl. microbiology)’. In summary, there is no robust and homogeneous level for data aggregation defined in MedDRA. This has also been recognised by Kübler et al.[4] and Kubota.[6]

Moreover, the HLT ‘Central nervous system haemorrhages and cerebrovascular accidents’ groups 53 PTs with opposite concepts such as ‘ischaemic stroke’ and ‘haemorrhagic stroke and bleeding’. The mix of opposite concepts in one HLT limits the value of this HLT for data aggregation. Therefore, such HLTs should be split into separate terms reflecting a single concept to allow for meaningful aggregation of cases. A systematic review of the HLTs by the MSSO is recommended to make the HLT level more usable for data analysis. Systematic review has also been suggested by Brown.[10]

It can be considered a weakness of MedDRA to sometimes mislead the user by its substructure, because, in some instances, a comprehensive grouping of a medical concept is suggested, such as the HLGT ‘allergic conditions’, but in fact only a subset of allergy-related terms can be identified under this HLGT. This is partly due to the fact that allergy is induced by immunological mechanisms and the primary SOC is, therefore, ‘Immune disorders’. However, most of the leading symptoms of allergy, in general and in this data set, are allocated to other disorder SOCs, such as ‘Skin and subcutaneous tissue disorders’ or ‘Respiratory, thoracic and mediastinal disorders’. Similar pitfalls are described in Brown.[10] It is hoped that these difficulties, which are not specific to MedDRA and exist in various degrees in all other terminologies[11] used for AE reporting, will be overcome in a standardised way by using well defined SMQs in the future, which are under development by a CIOMS/MSSO working group. The problem itself has been well known for a long time and has led to organisation-specific solutions, which should be overcome with a standard terminology.[11,12]

Multiaxiality

The rules related to multiaxiality can lead to situations where medically related concepts are assigned to different SOCs. If one wishes to aggregate data belonging to the same medical condition, it can be necessary to group terms from more than one SOC. In particular, the SOC ‘Investigations’ causes many concerns, because the fact that relevant laboratory findings are not linked to the disorder any more can lead to underestimation of risks.[4–6,12] Thus, the impact of multiaxiality on frequencies needs to be taken into account. This has been demonstrated in the example presented in section 2.2.

As a consequence of the multiaxiality of MedDRA and the rule that some SOCs are ‘stand alone’, it can occur that related medical concepts are represented under different SOCs. The HLGT ‘Hepatic and hepatobiliary disorders’ is under the SOC ‘Hepatobiliary disorders’ and the HLGT ‘Hepatobiliary investigations’ is under the SOC ‘Investigations’. However, to present all ADRs related to the liver, one would wish to group the related PTs under one category. It is important to realise that the reporting rate of a certain disorder may easily be underestimated if the laboratory and investigation terms classified under the SOC ‘Investigations’ are not systematically included in an analysis. In conclusion, the concept of organising abnormal laboratory findings under the stand alone SOC ‘Investigations’ makes it more difficult to determine the total number of cases belonging to a medical condition.

Similarly, while it can be very helpful for some searches to have all infections grouped under the primary SOC ‘Infections and infestations’, for the display of the overall safety profile of a drug this is of limited value because in a standard display by SOC it is not clear which kind of infections have been reported and to which part of the body they belong. The data presented under section 2.2 come from an antibacterial, which very rarely can cause pseudomembraneous colitis. This form of colitis is caused by an infection with clostridium difficile. According to the rules of MedDRA, the PT colitis pseudomembraneous is primarily assigned to the SOC ‘Gastrointestinal disorders’, but the PT clostridium colitis maps to the SOC ‘Infections and infestations’. The MedDRA rules say that all infections have to be displayed under the latter SOC as primary allocation. Obviously, one wishes to have both PTs grouped together for display of the ADR profile because they describe the same ADR. Such grouping is important in order to retrieve all related cases, because in some reports the colitis is mentioned and not the infection or the infectious agent and vice versa. It is also obvious that for the evaluation of the frequency of diarrhoea/colitis/pseudomembranous colitis, it is important to have a correct grouping of terms. By accepting the split into more than one SOC, one may underestimate the frequency of the medical condition under evaluation.

SSCs

Because of above-mentioned difficulties for data retrieval caused by current heterogeneity and conceptual ambiguity on some MedDRA group terms, analysis by medical concepts is only partly possible, and frequently a satisfactory grouping of terms under medical perspective would require SSCs. In the absence of a robust level for data aggregation in MedDRA by medical concepts, companies may be in the situation of creating their own search categories; there is a growing need for data analysis given that more and more companies have started coding their clinical projects in MedDRA. Issue analysis of spontaneous data is a permanent need of ongoing safety monitoring. However, creating organisation-specific search categories in order to overcome difficulties with a terminology is clearly against the MedDRA’s intention of having a common standard terminology. Also, developing organisation-specific solutions is very demanding in terms of resources and will create major inconsistencies between organisations. On the other hand, it is fair to recognise that these difficulties have also always existed with previous terminologies, but they are possibly more obvious with MedDRA.

While some of the current SSCs are focused on a clear concept, such as ‘Bone marrow depression’ or ‘Cardiac arrest’, others include extremely heterogeneous concepts, such as the SSC ‘Oedema’, which includes PTs such as ‘application site swelling’, ‘catheter site oedema’, ‘eye swelling’, ‘cerebral oedema’, ‘gingival swelling’, ‘laryngeal oedema’, ‘retinal oedema’ or ‘vaginal swelling’, etc. It is difficult to understand for what kind of searches such a mix of terms should be used.

It is hoped that SMQs will have a clear indication for use and a clear focus on a well defined medical concept; however, it is premature to judge on their actual utility, because only two of them have been published for review and testing.

Home-Grown Solutions versus Further Developing MedDRA

The approach used to combine existing elements in MedDRA by medical concepts presented under section 2.2 deviates from standard displays. It does group existing structural elements in a predefined way in an attempt to group the data by medical concepts that have been selected by the evaluator as the most relevant groupings for this drug. It does not change the structure or the links in MedDRA. The problem with this method is that, to a certain extent, it may be judged as arbitrary, and there is a need to clearly describe the rationale behind it. The situation may be remedied by linking the respective disorders’ SOC with the HLGT/HLT under the SOC ‘Investigations’, which can clearly be assigned to a disorder SOC. Under the SOC ‘Investigations’, only terms that cannot be linked to an organ system or that describe normal values or test names not relevant for safety reporting would remain. This assignment of investigation terms to disorder SOC should be done by the MSSO to guarantee a consistent approach.

The result displayed in figure 7 shows a significant shift in the numbers of PTs per SOC, for example, the percentage of terms in the SOC ‘Gastrointestinal disorders’ increased from 20% in the standard display by primary SOC to 32.34%, and the SOC ‘Hepatobiliary disorders’ reached 6.26% instead of 4% in the standard display. It was also striking that the SOC ‘Infections and infestations’ almost disappeared, whereas in the standard display it counted for 14.52%. Other shifts were less dramatic. Overall, the authors believe that such modified grouping of structural elements in MedDRA can add value to displays of ADR profiles, if the principles are transparently described.

However, such an approach is far from ideal because it deviates from standard approaches. The reasons behind this are the deficits of the current MedDRA, with regard to data aggregation (as described in section 3.4). One conclusion is that MedDRA in the current version has limitions for data analysis, because the PT level is too detailed and some group terms may need redesign to become a robust, consistent and non-ambiguous level for data aggregation. This redesign should be done as an interdisciplinary approach with medics and bio-statisticians and should take into account the needs of companies to base their product information on the results of data analysis of MedDRA-encoded data.

Summary

The structure and the rules and conventions of MedDRA have significant impact on the way data are actually coded. Simultaneously, data retrieval strategies, analysis, sorting and presentation of the data need to be reviewed and adapted according to the structure, rules and conventions of MedDRA. While it is recognised that a certain loss of information is unavoidable when narrative information is coded, MedDRA enhances the accuracy of coding by its richness of specific terms, which also helps to avoid interpretation at data entry. On the other hand, the high granularity of MedDRA can have a significant impact on data analysis, for example, on frequency tables. MedDRA’s structure also offers new options for graphical data presentation. However, MedDRA would benefit from a redesign particularly of the HLT level to become a robust, consistent and non-ambiguous level for data aggregation. This should be done as an interdisciplinary effort coordinated by the MSSO.

Conclusion

In summary, the authors come to the following conclusions.

  • MedDRA is different from any previously used terminology because of its complexity, specificity, granularity and structure.

  • The specificity of MedDRA improves accuracy and precision of data entry, particularly if consistency and accuracy in data entry are controlled by company guidelines and appropriate quality assurance procedures.

  • MedDRA in its current version has limitations for data analysis, mainly because the PT level is generally too granular and the group terms are not a robust, consistent and non-ambiguous level for data aggregation. The rules relating to multiaxiality add complexity for data aggregation because related medical concepts may be split.

  • In the absence of guidance from regulatory agencies and with the growing need for analysis of MedDRA-encoded data, there is a risk of diminishing the benefits of MedDRA as a standard terminology through the use of user-specific and non-standardised data output strategies.

  • MedDRA would benefit from a redesign particularly of the HLT level to become a robust, consistent and non-ambiguous level for data aggregation. This should be done in an interdisciplinary effort coordinated by the MSSO.